By Bill Honaker, The IP Guy
I’ve been intrigued by artificial intelligence ever since Steven Thaler first attempted to patent a couple of inventions that he claimed were created entirely by a creative machine, which he called DABAS. DABAS stands for, “Device for the Autonomous Bootstrapping of Unified Science.” He also tried to get copyright protection for a work of art called, “A Recent Entrance to Paradise,” that he claimed was entirely created by his creative machine. Both applications were denied protection because the law is clear. In the US, only humans can get patent or copyright protection. And since humans weren’t involved in the creation of these works, they could not be protected. Of course, there’s ongoing fights over whether that position is correct or not, and only time will tell if that viewpoint is going to change.
Then ChatGPT came out. A friend contacted me. He was playing with it and asked if I’d seen it, so I decided to give it a try. I prompted ChatGPT to write a 500 word article on the subject of intellectual property, and I was amazed. Within seconds I had this beautiful article, which was grammatically perfect, read extremely well, was very conversational and very authoritative.
But it was wrong, and it was subtly wrong, in very critical places. If you weren’t an expert in intellectual property law, you wouldn’t have even noticed that this very authoritative article was not giving you correct information. So since then, while watching AI develop, and particularly ChatGPT, I’ve learned about hallucination, and about how generative AI actually works.
To summarize — on a very simplistic level, because I’m not an expert in how AI works — AI is taking information from a database, supposedly the entire Internet, and it’s looking statistically at the next best word to pick, in order to answer the question you’ve asked in the prompt. So it’s not thinking, even though you’re led to believe it’s thinking, because they’re called “deep thinking machines.” It really isn’t thinking. It’s just statistically looking at what’s the next best word to use. If it can’t find the answer to your question, it will construct one. And it follows a template that presents that information in a form that’s most acceptable to humans. So when you read it, the grammar is correct, it’s authoritative, it’s conversational, it’s engaging, and all of those things make the article very interesting. But not necessarily trustworthy. And that’s part of the reason I don’t think generative AI is actually going to take our jobs. People who practice law, along with accountants, journalists, authors, and artists, all of those require human thinking, human input.
And what makes me even more intrigued by this prophecy is the recent revelation that these machines can go MAD. In this case, MAD is an acronym for, “Model Autophagy Disorder.” This occurs when generative AI consumes its own misinformation. I’m guessing the name is a reference to mad cow disease, which was transmitted by feeding cows the brains of other cows. Pundits are theorizing that within the next few years, almost 90% of the content on the Internet will be written by generative AI models like ChatGPT. So more and more, they will be consuming their own content, to the point of believing their own garbage and going MAD in the process. So with that in mind, the more I learn, the less I’m concerned.
This article is one of a series exploring the intellectual property aspects of AI, and how to protect AI-generated content. What can be protected, and what can’t, is within the big gray area we still live with. And I’m seeing, both at the copyright office and in the courts, some serious issues that we’re going to have to deal with as AI continues to develop.
AI is not new. In many respects, we’ve been dealing with AI for decades. It’s here to stay. It’s certainly going to develop further, but I don’t believe it’s going to take my job, or yours, any time soon.
Bill Honaker, “The IP Guy,” is a former USPTO Examiner, a partner with Dickinson-Wright, and author of the forthcoming book, Invisible Assets – How to Maximize the Hidden Value in Your Business. To download a sample chapter click here.